Researchers from the USC Viterbi School of Engineering and the School of Advanced Computing have created artificial neurons that closely mimic the complex electrochemical behavior of real brain cells. Their breakthrough, described in Nature Electronics, represents a major step forward in neuromorphic computing. This new approach could dramatically shrink chip size, cut energy use by several orders of magnitude, and bring us closer to achieving artificial general intelligence (AGI).
Unlike standard digital processors or existing silicon-based neuromorphic chips that only simulate neural activity, these artificial neurons physically reproduce the analog processes of biological neurons. In the same way that neurochemicals trigger brain activity, specific chemicals can now be used to initiate computation in brain-inspired, or neuromorphic, hardware. Because they replicate the actual biological mechanisms rather than relying on mathematical models, these artificial neurons are fundamentally different from earlier designs.
The research, led by USC Computer and Electrical Engineering Professor Joshua Yang—who previously made pioneering contributions to the field of artificial synapses—introduces a new type of artificial neuron built using what is known as a “diffusive memristor.” The study details how this innovation could enable a new generation of chips that enhance and extend today’s silicon-based technologies. While conventional electronics depend on the flow of electrons for computation, Yang’s diffusive device instead uses the movement of atoms. This atomic-level process allows the neurons to operate more like those in the human brain, offering greater energy efficiency and the potential to advance the development of AGI.
To read more, click here.